I made a change to the server tonight. Unless you're specifying QQQsecondpass then the server will hand out whatever k is lowest between error-fix, dropped-tests, and first-pass. This should focus us back on first-pass more.
The plan is eventually to get the priorities straightened out so we'll assign second-pass tests to keep second-pass not too far behind first-pass.
Apologies that it took so long for this to happen.. and I know some of you e-mailed me about it. I wasn't ignoring you on purpose, I was waiting for a response on some things which took a while...
Hurray!!
So, we are back on the right track. I hope the users who were not patient to do DC work will come back, and this project will stand it's old glory soon.Tell everybody the good news, everyone!
My apologies to have been annoying. You rule.
Yours H.
___________________________________________________________________
Sievers of all projects unite! You have nothing to lose but some PRP-residues.
I am happy to hear this .. thank you
Just a side note...
May be it's a good time to change the scale of the Pending Test distribution charts from 3m-10m to, say 5m-12m (15m?)
Mike,Originally Posted by Alien88
It looks like you now have the priorities sorted since the first-pass and error-fix are now being handed out in preference to double-check. Great job.
Out of interest, what are the ratios now set to?
I think there is no ratio ATM:
error-fix, dropped-tests and first-pass have the same priority, as seen on the "secret statistics" page.
It was, but it changed in the last 24 hours. I now have 21181•2^8640428+1, which is significantly higher than the lowest in the second-pass queue.
EDIT: Actuallly what I'm saing is that until 24ish hours ago second-pass was still at the same priority as the others.
thats no good.. 24737 still has tests < 3.5M and most still have some < 5M
but in general i agree with you.
It looks like, we'll see the tests n>10m as a vertical line until the upperhand of the scale is changed to a higher figure. Currently (at 10004000), it's only 2% of the pending tests, and it'll probably be more obvious within a couple of days as new tests are assinged.
It looks to me as though everything is similar to the way things were before. I guess we will have to wait for an error before we find out if it's assigned at secondpass or firstpass priority.
I think we will also always have the issue of small-n tests never been completed unless physical efforts are made.
I'm still puzzled by the garbage que however. I wonder if tests in there are acutally thirdpass now....
Regardless I think this is a moot point since you have to use garbage in order to access them.
------------------
As for a ratio between first and secondpass... I think the best bet is to simply change priorities manually. Run firstpass out to 12M or a prime... then switch everyone project wise to secondpass until 7M or 8M.
Sure the ratio will never be perfect but it will never get that far out of wack either.
--------------
As for the window I'm with Nuri let the small tests drop off the scale the way the larger tests spike the RHS. I'd say change the window to 5M-12M or even 6-12M.
Also the smallest out standing test is actually
n=1742073 for prime 27653
And
n=2717182 for 22699
Does this project actually run on a P60 (o/c'ed to 2x33=66Mhz SMMMMOOOOKE'IN!!! ), I know I have one but I'm not sure if it works anymore.
What about a short announcement on the home page? Now, it looks like we're still doublechecking...
H.
___________________________________________________________________
Sievers of all projects unite! You have nothing to lose but some PRP-residues.
Nice catch..
hhh,
I know there are alot of people from ARS who simply changed their machines over to QQQsecondpass when the firstpass que ran empty. Then never changed the username back during the whole project wide secondpass effort.
If you remember back when there was a period of a couple days when this was the only way to get a test, so I think there may be quite a few people who did this. I wonder how many people are still using secondpass in thier username?
Personally I don't think this is a bad idea to keep a few people and slower machines there... I think people are a little less worried about running that que now b/c the scoring isn't as bad as it use to be.
As for a short anouncement? Do you really think we need to try to change the ratio of people? I'd personally leave things alone for a while. We are doing roughly the same number of tests in each que and secondpass is still a valid effort.
Given that the n increase on the second-pass queue is currently +864, which equates to 18 tests being issued in the last 24 hours, there can't be too many folks with PCs set to QQQsec.. Contrast that with the first-pass n increase of +12094, which equates to fairly healthy 257 tests being issued
I think maybe there is a feeling that users deserted in droves when we moved to second pass, and unless they look on this forum they'll have no idea that normal service has resumed.
Yup, that's what I meant. Sorry not to have been precise enough. And even new users might be frightened by the idea of doing doublecheck.I think maybe there is a feeling that users deserted in droves when we moved to second pass, and unless they look on this forum they'll have no idea that normal service has resumed.
___________________________________________________________________
Sievers of all projects unite! You have nothing to lose but some PRP-residues.
Humm... I always look at the results being returned as opposed to those handed out. But your right most people are firstpassing now.
On the annoucment, very good point, besides is there any harm in an annoucement that we have started testing n>10M... it is a milestone.
Looking at the data in regards to user loss...
http://www.seventeenorbust.com/stats/
I believe we started secondpass testing project wide around late ?sept?, the rate was approximately 1900M then. Currently we could guess that the rate is around 1750M, so droves is probably not the case but loss (it appears so).
Our overall rate decreased and we have not seen an increase due to faster processors, so how many people left? 20% of the people... sounds high?
But looking at the number of people active in the last 2 weeks it's 930, this number use to be above 1000.
???? Pray for a prime ???
May be not so... May be, we're a bit dillusioned due to the fact that still roughly half of the pending tests are second pass tests, and the case is not that bad. It might be a good idea to wait for a couple of days before reaching any conclusions.
Nuri, you may in fact be correct since there are the few people who don't submit intermediate blocks. In addition I'm not sure if the daily rate is a true daily rate or if its averaged over the last few days etc...
I'm sure by end of feb we will really know when the true rate is at. Regardless I think hhh and MikeH are correct an anoucment should be made.
I already made an annoucement at ARS begining of last week.
There is also quite a bit of buzz lately about bionic and other projects closing, SETI, Folding etc,... I think people may be trying these project just for a change. They may come back we may get others...
What would really help would be a true 64-bit version of the program... Since I don't have a 64-bit processor I havn't been following the latest clients. Is there a 64-bit factoring version of Prime95?
I think this may be an issue RC-72 has a 64-bit, couple other projects as well. Now if they are truly 64-bit is another question and are there advantages over 32-bit... if we had a 64-bit version of anything we would have an advantage. In a few months we will need a 64-bit client to compete
Many thanks Mike.
I noticed that we are still getting alot of tests complete in the error fix... I'm pretty sure these will go away in a few days but I think it's something we need to keep an eye on.
I never send intermediate blocks which always makes my stats interesting.
http://www.seventeenorbust.com/stats...ml?userID=4841
My last day's rate is always low but my Overall rate keeps going up.
As of today, more than half of the pending tests are above 10m.
Posting Permissions
Reply With Quote